Paste text or simulate a size, and see how much of each model's memory it fills. No login, nothing saved.
How far this fills each model's context window
It's one shared pool
Your instructions, any files you attached, the whole conversation history, and the AI's reply all draw from the same limited memory — not separate allowances for each.
More isn't always better
Dumping a huge document in often gets a worse answer than a short, focused excerpt. AI tools tend to skim over details buried in the middle of long content.
It's why long chats get worse
The longer a conversation runs, the more likely it is to start forgetting earlier details. That's this same limit being reached, not the AI "getting tired."
LLM context windows, explained
What a context window is and why bigger isn't always better.
What happens when you hit the limit
Claude Code's context window behavior, and how to fix it.
Token budget: planning & execution
How to plan a token budget for agents and long workflows.
Try the Tokenizer Playground →
Watch text split into tokens live, token by token.
It's an AI model's short-term memory for one conversation — measured in tokens, not words. Your message, any files you've shared, everything said earlier in the chat, and the AI's own reply all draw from that same limited pool. Fill it up, and something has to give.
It depends on the model, and providers update these numbers often. As a rough snapshot: Claude's Sonnet, Opus, and Haiku models hold around 200,000 tokens (roughly a 500-page book). Some Google, OpenAI, and Meta models go much higher, into the millions. Bigger doesn't always mean better — see the next question.
Not necessarily. AI models tend to pay less attention to details buried in the middle of a very long conversation — the same way you'd remember the start and end of a long meeting better than the middle. A short, focused message often gets a better answer than a giant wall of pasted text.
Depending on the tool, one of a few things happens: it stops you with an error, it quietly drops or summarizes your earliest messages, or it starts "forgetting" details from earlier in the conversation. If an AI chat suddenly seems to lose the thread, this is usually why.
It's a close estimate, close enough to get the idea. Different AI companies count tokens slightly differently and don't all publish their exact method, so actual counts can vary by roughly 10-20% between tools for the same text.
Tokenizer Playground
Type anything and watch it split into tokens live — the units LLMs actually read and get billed for.
FROG in a Bowl Prompt Builder
Fill in Format, Role, Objective, Goal, and Context — get a copy-ready structured prompt in seconds.
Machine Learning Types
Three tiny games show what supervised, unsupervised, and reinforcement learning actually mean.
Neural Network Playground
Drag two sliders and watch a real, tiny neural network turn them into a decision, live.
Generative AI Playground
Watch AI write one word at a time by predicting what's most likely to come next.
RAG Playground
Ask a question, watch notes get retrieved, then see a grounded answer versus a hallucination.
Embedding Map
Click two words and see why similar meanings sit near each other — the idea behind vector search.
Attention Visualizer
Click a word and see which others a toy transformer looks at — including the classic “it” puzzle.
Prompt Injection Lab
Watch a pasted email try to override a support agent, then flip a switch that treats it as data.
